AI Provider Privacy

Privacy implications of different AI providers in Dayflow

Dayflow lets you choose your AI provider, and each has different privacy implications. This page helps you understand what data leaves your Mac with each option.

Privacy Comparison

Local Models

Fully private - Nothing leaves your Mac

Gemini

Cloud processing - Configure data retention

ChatGPT / Claude

Third-party processing - Review their policies

Provider Details

Local Models (Ollama / LM Studio)

Privacy level: Highest

What Stays on Your Mac

• Everything: All processing happens locally
• Screen recordings never leave your machine
• AI models run entirely on-device
• No internet connection required for processing

Privacy Benefits

• Zero cloud exposure: Your screen data never touches external servers
• Offline capable: Works without internet (after models are downloaded)
• Full control: Inspect models, control when they run

Trade-offs

• Quality: Local models may underperform cloud models on complex summarization
• Speed: Slower than cloud-based processing, especially on older Macs
• Battery: GPU-intensive processing drains battery faster (recommend plugged-in use)
• Setup: Requires installing and configuring Ollama or LM Studio

Recommended for: Users who prioritize maximum privacy and don’t mind trade-offs in quality/speed.

References

• LM Studio offline documentation
• Ollama GPU acceleration

---

Gemini (Google AI)

Privacy level: Configurable

What Leaves Your Mac

• Screen recording chunks: Uploaded to Google’s Gemini API for analysis
• Batch payloads: Sent periodically during timeline generation
• No persistent storage by Dayflow: Data is sent for processing only

Data Handling: Paid vs. Unpaid Services

Google’s data handling depends on whether you’ve enabled Cloud Billing:

With Cloud Billing (Recommended)

Paid Services Data Handling

When you enable Cloud Billing on at least one Gemini API project:

All of your Gemini API and Google AI Studio usage is treated under “Paid Services” rules - even when using free quota.

What this means:

• Google does not use your prompts/responses to improve Google products or models
• Your data is not used for training
• Applies to all Gemini API calls, not just billed ones

Important nuances:

• AI Studio remains free to use - enabling billing only changes data handling, not pricing
• Abuse monitoring: Google logs prompts/responses for a limited period for policy enforcement and legal compliance
• Verification: In AI Studio → API keys, you should see “Plan: Paid” once billing is enabled

From Gemini API Additional Terms:

“When you activate a Cloud Billing account, all use of Gemini API and Google AI Studio is a ‘Paid Service’ with respect to how Google Uses Your Data, even when using Services that are offered free of charge.”

Without Cloud Billing

Unpaid Services Data Handling

Without Cloud Billing enabled:

Google may use your prompts and responses to improve Google products and services, including training models.

Exception: EEA / UK / Switzerland users

If you’re in the European Economic Area, UK, or Switzerland, the Paid Services data handling applies by default - even without enabling billing.

This means:

• Your data is not used for training (even on free tier)
• Same privacy protections as Paid Services
• No billing configuration required

From Gemini API Additional Terms:

“For users in the European Economic Area, Switzerland, and the United Kingdom, the Paid Services data handling described above applies to all Services (including AI Studio and free Services).”

How to Enable Cloud Billing

1. Go to Google Cloud Console
2. Select (or create) a project
3. Navigate to Billing
4. Enable billing for the project (you won’t be charged unless you exceed free tier limits)
5. Verify in AI Studio that your plan shows as “Paid”

You can enable billing on one project and continue using API keys from any project - the data handling applies account-wide once any billing is activated.

Trade-offs

• Quality: Excellent - native video understanding for direct analysis
• Speed: Fast - only 2 LLM calls per recording chunk
• Efficiency: Most efficient provider (fewest API calls)
• Privacy: Depends on billing configuration (see above)

Important: This is our interpretation of Google’s Terms of Service, not legal advice. Review the full terms yourself.

---

ChatGPT / Claude (CLI-based)

Privacy level: Third-party processing

What Leaves Your Mac

• Screen recording chunks: Processed by OpenAI (ChatGPT) or Anthropic (Claude)
• Extracted frames: Sent via CLI tools that communicate with cloud APIs
• Batch processing: Frames extracted every 60s, sent in batches

Data Handling

Your data is processed by third-party AI providers:

ChatGPT (OpenAI)

Provider: OpenAI

How it works:

• Dayflow uses the Codex CLI to send frames to OpenAI
• Requires ChatGPT Plus ($20/month) or ChatGPT Pro subscription
• CLI authenticates through your existing ChatGPT account

Privacy policy:

• Review OpenAI’s Privacy Policy
• Enterprise users: Consider ChatGPT Enterprise for enhanced data controls

Data usage:

• OpenAI’s policy governs how your data is used
• As of 2024, OpenAI states they don’t train on ChatGPT Plus/Pro user data by default
• Verify current terms on OpenAI’s website

Claude (Anthropic)

Provider: Anthropic

How it works:

• Dayflow uses Claude Code CLI to send frames to Anthropic
• Requires Claude Pro subscription ($20/month)
• CLI authenticates through your existing Claude account

Privacy policy:

• Review Anthropic’s Privacy Policy
• Anthropic’s Commercial Terms may apply depending on your use

Data usage:

• Anthropic’s policy governs how your data is used
• Review their current data retention and training policies

Trade-offs

• Quality: Highest - frontier reasoning models provide best-in-class narratives
• Speed: Good - 4-6 LLM calls per recording chunk
• Cost: Requires paid subscription ($20/month)
• Setup: Requires CLI installation and staying signed in
• Internet: Requires active internet connection (no offline mode)

When using ChatGPT or Claude, you’re subject to their privacy policies, not just Dayflow’s. Review their terms carefully.

---

Quick Comparison Table

Feature	Local Models	Gemini	ChatGPT / Claude
Data leaves Mac	No	Yes	Yes
Cloud processing	No	Yes	Yes
Training on your data	N/A	No (with billing)	Check provider policy
Offline capable	Yes (after model download)	No	No
Internet required	Only for model downloads	Yes	Yes
Cost	Free (after hardware)	API costs (free tier available)	$20/month subscription
Quality	Good	Excellent	Highest
Speed	Slower	Fast	Good
Battery impact	High (GPU-intensive)	Low	Low
Setup complexity	Medium	Easy	Medium
LLM calls per chunk	33+	2	4-6

Choosing the Right Provider

Choose Local Models if:

• Privacy is your top priority
• You work with highly sensitive information
• You want offline capability
• You don’t mind slower processing and lower quality
• You have a Mac with sufficient GPU power

Choose Gemini if:

• You want the best balance of quality, speed, and cost
• You’re comfortable with Google’s data handling (especially with billing enabled)
• You want the most efficient processing (fewest API calls)
• You’re in EEA/UK/Switzerland (automatic Paid Services treatment)

Choose ChatGPT / Claude if:

• You want the absolute highest quality narratives
• You already have a ChatGPT Plus/Pro or Claude Pro subscription
• You trust OpenAI’s or Anthropic’s data handling
• You need best-in-class summarization quality

Switching Providers

You can change AI providers at any time:

1. Open Dayflow Settings
2. Go to AI Provider
3. Select your preferred provider
4. Configure API keys or CLI tools as needed

Previous timeline data remains intact - only future processing uses the new provider.

Next Steps

Privacy Overview

Understand Dayflow’s privacy architecture

Data Storage

Learn where your data is stored locally